You are on page 1of 22

# Quantitative Methods 2010

Probability Theory

QM 2010 - Kingston

Overview
Event, experiment, and sample space Addition rule for probabilities Multiplication rule for probabilities Bayes Theorem

QM 2010 - Kingston

## Event, Experiment and Sample Space

Event
An outcome from doing something Two or more events can be clubbed together and spoken of as a single event

Experiment
The activity that produces an event

Sample space
The set of all possible outcomes of an experiment
QM 2010 - Kingston 3

Probability
Let A be some event. Based on prior knowledge (classical probability) No. of ways A can happen P[ A] Tot no. of all possible events Based on observations in the past (relative frequency of occurrence)
No. of times A happened P[ A] Tot no. of all possible events Based on subjective estimation

Note: When only 1 event can take place, above probabilities are known as marginal probabilities or un-conditional probabilities
QM 2010 - Kingston 4

## Examples Classical Probability

Let P*A+ be P*Getting head in 1 toss+ *
1 p[ A] 2
No of ways of getting head Tot no of all possible outcomes

## Let P[E] be P[fresher entering wrong LH] **

7 p[ E ] 8
No of wrong LH Tot no of all LH ** Assuming he is told to go only on 1st floor LH

## * Assuming coin is fair

QM 2010 - Kingston

## Examples Past Observations

Let P[B] be P[milk-pouch is bad+
In batches of 200 pouches received in the past, average 3 were found to be bad
3 p[ B] 0.015 1.5% 200

## Let P*x+ be P* companys IPO will be oversubscribed]

Out of 350 IPOs so far this year, 210 were oversubscribed
210 p[ X ] 0.6 60% 350
QM 2010 - Kingston 6

Venn Diagrams
Entire sample space is represented by a rectangle The set of events we are interested in is represented by any closed curve (circle, ellipse, square, etc). If 2 or more events are mutually exclusive, their closed curves are shown not overlapping.
If not (if they can occur together), their closed curves are shown overlapping.

Parts from supplier A Parts from supplier B QM 2010 Parts from all suppliers (sample space)Kingston

From supplier A (sample space) Parts with wrong length 7 Parts with wrong diameter Parts with wrong length AND wrong diameter

(Statistical) Independence
Two events (and their respective probabilities) are said to be (statistically) independent, if the probability of either of them happening does not affect the probability of the other happening. Example
Two suppliers, A and B, have supplied batch of 1000 parts each, of which some parts are known to be bad (defective). Let P[A] be P[picking bad part from A], and P[B] be P[picking bad part from B] Then above events and the respective probabilities P[A] and P[B] are independent

QM 2010 - Kingston

(Statistical) Dependence
Two events (and their respective probabilities) are said to be (statistically) dependent, if the probability of either of them happening DOES affect the probability of the other happening. Example
Consider any 1 supplier, A. Let P[A,1] be P[1st part picked is bad, and P[A,2] be P[2nd part picked is bad]. Then above events, and their respective probabilities are dependent.
QM 2010 - Kingston 9

## AND Probabilities (Joint Probabilities)

When two events are of interest, we may want to know the probability that both can happen together

## Joint probability of 2 events A & B is written P[AB] If A & B are independent

P[AB] = P[A]*P[B]

## If A & B are dependent

P[AB] = P[A]*P[B/A]
QM 2010 - Kingston

10

## Examples For AND Probabilities

Example 1
Five families A,B, C, D, E are waiting for a flat allotment from MHADA. Only 1 flat is available, and family will be chosen by random draw. But 2 of the families are retired Navy families, which 2 is not known
What is the probability that flat will be given to family D, AND to a Navy family?
P[D is chosen] = 1/5 = 0.2, P[Navy family is chosen] = 2/5 = 0.4 P[flat to D AND to Navy family] = 0.2*0.4 = 0.08
QM 2010 - Kingston 11

8%

20 %

40 %

## P[D is chosen] P[Navy is chosen] P[someone is chosen] (sample space)

QM 2010 - Kingston

12

OR Probabilities
When two events can happen, we may want to know the probability of either one OR the other happening OR probability of events A & B is written P[A +B] or sometimes P[A or B] If events are mutually exclusive
P[A or B] = P[A] + P[B]

## If events are not mutually exclusive

P[A or B] = P[A] + P[B] P[AB]
QM 2010 - Kingston 13

## Venn Diagrams For OR Probabilities

Mutually exclusive events Not mutually exclusive events

Parts from supplier A Parts from supplier B Parts from all suppliers (sample space)

From supplier A (sample space) Parts with wrong length Parts with wrong diameter Parts with wrong length AND wrong diameter

In straight sum, overlapping area Required probability is straight sum would be counted twice. So required of the 2 areas probability is straight sum of the 2 QM 2010 - Kingston minus the overlapping area. areas,

14

## Examples for OR Probabilities

Example 1
Five families A, B, C, D, E are waiting for a flat allotment by MHADA. Only 1 flat is available, and family will be chosen by random draw. What is the probability that family B or D is chosen?
P[B or D is chosen] = P[B is chosen] + P[D is chosen] = 1/5 + 1/5 = 2/5 = 0.4 = 40%

QM 2010 - Kingston

15

20 %

20 %

## P[D is chosen] P[B is chosen] P[someone is chosen] (sample space)

QM 2010 - Kingston

16

## Examples for OR Probabilities

Example 2
Five families A, B, C, D, , E are waiting for a flat allotment from MHADA. Only 1 flat is available, and family will be chosen by random draw. But 2 of the families are retired Navy families, which 2 is not known
What is the probability that family D is chosen or a Navy family is chosen?
P[D is chosen] = 1/5 = 0.2, P[Navy family is chosen] = 2/5 = 0.4, P[D is Navy family] = P[D]*P[Navy Family] = 0.2*0.4 = 0.08 Required probability = P[D] + P[Navy family] P[D is Navy family] = 0.2 + 0.4 0.08 = 0.52

QM 2010 - Kingston

17

8%

20 %

40 %

## P[D is chosen] P[Navy is chosen] P[someone is chosen] (sample space)

QM 2010 - Kingston

18

Bayes Theorem
Useful in situations where you know the probability of an event (its prior probability) , but you need a revised probability (its posterior probability), given that some other event has occurred. Example
You know the probability of winning a contract. You would like to know what is the probability of winning, given that one of your competitors bid has been rejected.

## P[B/A] = P[BA]/P[A] --- Bayes Theorem

QM 2010 - Kingston 19

## Bayes Theorem Scenario

Die 1 P[Sixer] = 40%, die 2 P[Sixer] = 70%
Pick a die at random, and it comes Sixer What is the probability that it is Die 1?

Prior probability of Die 1 = 0.5, Die 2 also = 0.5 For better answer we form following table:
Die Choice P[Die Choice] Die 1 Die 2 0.5 0.5 P[Sixer|Die Choice] 0.4 0.7 P[Sixer from 1 OR 2] =
QM 2010 - Kingston

20

## Bayes Theorem Scenario

So P[Die 1|Sixer] = P[Die1 AND Sixer]/P[Sixer]
= 0.20/0.55 = 0.364

Similarly, P[Die 2|Sixer] = 0.35/0.55 = 0.636 Conclusion: 1 toss result has allowed us to get better estimate of P[Die 1] (and Die 2). Application: Managers can make better decisions with Bayesian Decision theory using additional information.
QM 2010 - Kingston 21

## End Of Session Probability Theory

QM 2010 - Kingston

22